Facial recognition technology could soon tailor shop advertising

Developments in the field could mean that technology that has long been the stuff of science fiction could now be used in the real world, with views differing on whether that is a good thing or not.

Keith Raderschadt from NEC Corporation of America gives a detailed explanation and demonstration of the new facial recognition software being implemented by the Calgary Police Service at their CPS Headquarters, Westwinds Campus Media Centre in Calgary. Darren Makowichuk / QMI /Corbis
Powered by automated translation

It has long been the stuff of science fiction: entry systems that can recognise individuals from their face alone.

There has long been a fascination with the idea that technology can decide who to let in or keep out.

Facial recognition technology was a key development highlighted by organisers of the recent Hotel Show Dubai, held at the Dubai World Trade Centre.

There are now passport control gates that use biometrics to decide, without human input, whether individuals can pass through. Technology has the potential to combat terrorism.

“Governments are highlighting security issues and they’re putting funding into these areas,” says Tim Cootes, professor of computer vision at the University of Manchester in the UK, adding that the US authorities are pushing the sector forwards.

“The ability to identify an individual is considered a priority for security and the digital economy,” he said.

The digital economy is important because businesses are increasingly looking at how they can use facial recognition technology.

It could be used to identify particular individuals that are especially welcome, such as regular store customers or VIP hotel guests, to tailor advertisements to individuals by determining their age and gender, or to highlight undesirables, such as criminals, who should be kept out or monitored. Facial recognition technology involves using a digital video camera to capture images of a person’s face and then calculating values for a number of the face’s “nodal points”.

There are dozens of these, such as how wide the nose is, how far apart the eyes are and how deep the eye sockets are. When added together, they create a faceprint.

Typically, systems now build up a three-dimensional picture.

These systems “are getting reasonably good at verifying if somebody is who they say they are”, according to Prof Cootes.

But in the real world, things tend to be more problematic.

“If they are trying to be recognised, they can make it easy for the system. The systems are making errors of one per cent or less in the lab. As soon as you go outside, the systems are much less reliable because of variations in lighting and appearance,” he says.

“You wouldn’t want to use it for gaining access to your bank account – it would make far too many mistakes.” However, the technology is becoming more sophisticated and reliable, and better able to deal with complicating factors, such as changes in expression.

“If you know generally how people’s faces vary with expressions, you can recognise them without needing to have that person smiling or frowning in your database,” says Adrian Evans, head of the department of electronic and electrical engineering at the University of Bath, UK. Dr Evans is looking at how people can be identified from just their nose and the region surrounding it.

This is a particularly useful area to use for recognition because it tends to remain the same regardless of expression. Facial regions where bone and rigid tissues are visible are usually the most helpful for recognition. Facial recognition company Aurora, based in Northamptonshire, UK, has exported to the UAE, where its technology is used to clock workers in and out.

“Our system is used in quite a lot of construction sites in the Middle East, in Abu Dhabi and Dubai,” says chief executive Hugh Carr-Archer.

“Most people don’t mind showing their face, but people are resistant to fingerprints – it has connotations to criminality.”

When it comes to online uses of facial recognition technology, there have been advances, and it is easier to identify people even if images are of poor quality.

Still, systems remain far from perfect. In a widely publicised recent case, Google was “appalled and genuinely sorry” when facial recognition software used by its Google Photos system misidentified two black people as gorillas.

The company admitted that the same system had labelled white people as dogs or seals.

Other biometrics aside from the face can be used to improve reliability: fingerprints, fingernails and iris scans, among others. “If you have several biometrics together, the whole system gets stronger,” says Dr Evans.

The use of multiple biometrics makes it more difficult for people to fool facial recognition systems, although it is still possible, with as little as a pair of sunglasses or a beard to help prevent identification.

There are also concerns for the opposite reason – that the technology is becoming too effective at identifying people and that companies can or will be able to identify people even if they do not want this to happen. In the European Union, permission must be sought before facial recognition technology is used for commercial purposes, but elsewhere, especially in the United States, regulation is lighter.

Bella Sankey, director of policy at the civil liberties organisation Liberty, described facial recognition technology as a “deeply intrusive capability”.

“What’s sold to us in the name of convenience quickly becomes creepy when you consider the current ambitions of corporates and governments to hoover up our personal information and profile and predict our every move,” she said.

London-based Big Brother Watch also had concerns. Daniel Nesbitt, the organisation’s research director, said: “More companies are exploring the potential uses of facial recognition. However, it is vital before using it that they ensure the highest levels of protection are in place to safeguard people against any misuse by what can be a hugely invasive technology.”

Reports have also raised fears that companies or employers could use information obtained with the technology to deny individuals insurance cover or jobs. Some critics have suggested there will be a time when the moment a customer walks into a store, the staff will know who that person is and how much he or she earns. Dr Evans says, however, that firms will not be identifying individuals, and will instead focus on their broader characteristics, such as their age.

“It’s something to be concerned about but, at this stage, people probably need to look at the stuff [they make] available online freely,” he says.

Despite the concerns of campaigners, the technology is set to be used more widely. So, in future, when we are waiting to pay in a shop and a nearby advertising screen starts showing products that apply closely to someone of our age and gender, we’ll know why.


Daniel Bardsley is a UK-based freelance journalist and former reporter at The National. He has science degrees from the University of Oxford and the University of East Anglia.